"37 million people have kidney disease. It may not be the same disease for all 37 million people. And so if we can apply some of these principles in terms of genetics and functional genomics, maybe we could have better insight onto how to have better therapy."
Jason Coloma is a trained scientist who spent years in drug discovery at Genentech before the company was acquired by Roche, moved to Basel as part of the integration, eventually led business development for oncology and research technology platforms, and then made a deliberate exit into early-stage biotech formation at Third Rock Ventures. The ask to become CEO of Maze — the company he helped build — came about seven years ago. He said yes.
This conversation covers what Maze is actually doing: applying variant functionalization to kidney disease to generate mechanism-based therapies where none exist today. It covers APOL1-mediated kidney disease in depth — how it was discovered, what it does to kidney cells, who it affects, and why current standard-of-care therapies don't work for these patients. It also covers the infrastructure problem of building a genetically defined patient population in a disease area where precision medicine has historically been absent, and where AI genuinely helps in drug discovery versus where the hype outruns the reality.
From Genentech to Building Maze Therapeutics
Coloma's path into biotech entrepreneurship was deliberate but not linear. Coming out of big pharma — having observed Roche and Genentech from inside during and after a major acquisition — he had a clear sense of what it took to develop medicines, but less exposure to how small companies were built from nothing.
Third Rock gave him what he described as a "postdoc in entrepreneurship." He watched how they built companies — Myocardia (acquired by BMS), Revolution Medicine, Blueprint Medicine (acquired after reaching commercial stage) — before he was handed the opportunity to build Maze. One of his co-founders was Charles Hamsy, who had co-founded Myocardia and Bridge Bio alongside Neil Kumar. "To have people like that around who could really think about how do you apply some of the things we were trying to do from a genetic standpoint, but also keep in mind how to think about this for patients — having drug hunters like that around us at the beginning was tremendous."
The conviction to build came from a simple observation: precision medicine had transformed outcomes in oncology by matching patients to therapies based on oncogenic drivers. Nobody had seriously applied that logic to kidney disease.
Applying Precision Medicine Beyond Oncology
Kidney disease affects 37 million people in the United States and is now the ninth leading cause of death globally — a fact that surprises most people who hear it. Nephrologists describe their primary job as keeping patients out of dialysis for as long as possible, because once patients enter dialysis, outcomes deteriorate sharply. Coloma described a graph Maze uses to make the point: the 10-year survival curve for dialysis patients looks nearly identical to the survival curve for metastatic breast cancer.
"People don't really recognize that," he said. "Once you're in dialysis, your outcomes are horrible."
Despite the scale of the disease and the severity of end-stage outcomes, kidney disease had seen almost no precision medicine development. The drugs available were mostly supportive — ACE inhibitors, ARBs, immunosuppressants — designed to manage progression rather than address underlying cause. Coloma's thesis: if you could subsegment 37 million patients into genetically defined subpopulations, you might find diseases within the disease, each with a tractable mechanism and a targetable pathway.
Moving Beyond Statistical Genetic Associations
The problem with most genomic research in disease is that it stops at correlation. Genome-wide association studies (GWAS) identify statistical relationships between gene variants and disease outcomes. These associations get published, sometimes make the news, and rarely become medicines. "The probability that those actually transition from a genetic insight to a medicine concept is actually pretty low," Coloma said.
What Maze built instead was a variant functionalization platform — a set of experimental and computational tools for understanding not just whether a gene variant correlates with disease, but how it causes disease. The concept originated in papers out of the Broad Institute roughly 15 years ago, but at the time, it couldn't be operationalized: the data sets didn't exist, the tools weren't there.
What changed: the UK Biobank and other longitudinal genetic cohorts with hundreds of thousands to millions of individuals, paired genetic and clinical data tracked over time. CRISPR. Single-cell genomics. Cryo-EM for structural understanding of proteins. AlphaFold. "These things started to emerge — if you properly integrate them, you can move better from a genetic association to a drug candidate."
Maze built an integrated platform that runs every program through the same sequence: identify the genetic signal, functionalize the variants to understand mechanism, build a molecule informed by that mechanism, and use the genetics to identify the patients most likely to benefit.
APOL1 Mechanism Discovery
Maze's lead program is APOL1-mediated kidney disease — a condition driven by two specific genetic variants (called G1 and G2) that cause a toxic gain of function in the APOL1 protein. The statistical association had been in the literature for 10 to 15 years. No one had established the mechanism.
Maze used its platform to do two things. First, they identified a protective variant — N264K — by mining large data sets. People who carried both the disease-causing G1 or G2 variants and this protective variant were not developing kidney disease, or were progressing much more slowly. "That's interesting — they were protected from eventually forming kidney disease. We found that protective variant. It hadn't been published before."
They then brought that variant into the lab to study how it was protecting. The answer revealed the disease mechanism: APOL1 proteins, under certain conditions, overexpress in podocytes — the kidney cells that maintain the filtration membrane — and insert themselves directly into the cell membrane, punching holes in the kidney cells. Those holes have no gate. They allow damaging molecules into the cell, causing injury and cell death.
"Now we know what is actually causing the nephrotoxicity — the kidney injury. So let's try to make a drug that prevents that." The drug Maze built blocks the pore, preventing damaging molecules from entering the cell — and ideally prevents the pore from forming in the first place.
Clinical Reality of APOL1-Mediated Kidney Disease
APOL1-mediated kidney disease disproportionately affects Black communities. The G1 and G2 variants are found almost exclusively in people of West African ancestry. The clinical presentation is distinct from typical kidney disease in three important ways.
First, patients present earlier — typically before age 50, compared to much later in life for kidney disease without the genetic driver. Second, they progress faster — entering end-stage renal disease and dialysis approximately 10 years earlier than patients without the variant. Third, they don't respond to current standard-of-care therapies: ACE inhibitors, ARBs, and immunosuppressants produce little to no benefit in these patients.
The diagnostic journey typically starts when a younger patient comes in for an annual checkup with unexplained high blood pressure. A workup reveals elevated protein in the urine — UACR, a standard lab measurement. Normal is below 30; patients with APOL1 kidney disease can present with UACR values of 2,000 or above. They're referred to a nephrologist, monitored, given immunosuppressants, and if eGFR continues to fall and protein continues to rise, they progress to end-stage disease — with a four-year wait list for a kidney transplant as the downstream reality.
"All these therapies don't get to the underlying cause of disease," Coloma said. "Let's go after truly what might be causing disease — APOL1 itself."
Building Infrastructure for Genetic Testing
Applying precision medicine to kidney disease required building infrastructure that barely existed when Maze started. "You don't go into your primary care and say, get a battery of genetic tests — that's not how they're treated."
Three things had to develop in parallel. First, patient advocacy and awareness: organizations like Nephe and the Kidney Health Initiative have started making APOL1 kidney disease more visible and encouraging genotyping for patients with kidney disease of unknown cause. Academic working groups like Parasol have been engaging with the FDA on appropriate surrogate endpoints and accelerated approval pathways. "This stuff didn't exist 10 years ago."
Second, testing infrastructure: APOL1 genotyping is now available on standard LabCorp and Quest panels, which means nephrologists can order it without a specialist referral. Third, ICD-10 classification: a diagnostic code for APOL1-mediated kidney disease allows physicians to build testing workflows and eventually seek reimbursement — a practical prerequisite for widespread adoption.
"We've seen inroads in all of that. Still early. A lot of work has to be done. But what's promising is that even early in development, you're seeing this — as opposed to other diseases where it takes even longer to get to that level of infrastructure."
Authentic Community Engagement
Because APOL1 kidney disease predominantly affects the Black community, Maze has had to build community engagement infrastructure that is unusual for a mid-stage biotech. "In oncology development, we didn't need to do that. So this is a different game."
Maze has a patient advocacy team actively engaging not just nephrologists but broader community institutions. They've had conversations with community leaders — including a pastor of one of the largest congregations in Philadelphia — about what appropriate engagement looks like when you're in mid-stage development and can't make promises about outcomes.
"You don't want to overpromise. How do you authentically do that with people such that they also see that? And how do you build trust? Because it does take trust-building over time." The core message Maze brings to community conversations: help us ask this question together. There are no approved therapies. The only way forward is through development, through answering the science, and through making that process accessible to the people it's meant to serve.
Near-Term Milestones and Pipeline Progress
Maze has critical data coming on the APOL1 program — if the data reads out as hoped, the company plans to move into a registrational study, the final stage before seeking approval. "That would be tremendous if we can actually transition to that. That is the goal."
Beyond APOL1, Maze has a second program using the same genetics-driven platform that generated well-received data in September of the prior year; Phase 2 is expected to start in 2025. The pipeline reflects the platform thesis: the same machinery — variant functionalization, mechanism discovery, genetically defined patient selection — applied to multiple disease areas where the conditions are right.
Where Genetics-Driven Platforms Work — and Where They Don't
Coloma was direct about the limits of Maze's approach. It doesn't work for diseases where clinical biomarkers don't translate reliably to patient outcomes, and where the underlying pathophysiology isn't well understood.
Kidney disease is well-suited: protein in urine (UACR) and kidney filtration rate (eGFR) are quantitative, longitudinally trackable, and strongly correlated with outcomes. The same is true of certain cardiovascular diseases and metabolic syndrome.
Neurological diseases are a different story. "People ask us all the time — why don't you do this for Alzheimer's, why don't you do this for Parkinson's? The problem is the biomarkers — including some that have emerged like neurofilament light chain — don't really correlate to outcome. We don't know if measuring those markers is going to impact the disease state or even outcomes." The amyloid hypothesis in Alzheimer's is a case in point: studies that successfully reduced amyloid burden didn't change disease trajectory as expected. Without a reliable biomarker-outcome link and large enough longitudinal genetic data sets, variant functionalization doesn't produce clean signal.
"I think it might be at this point a little bit early and a little bit overused in diseases where you don't have that relationship." Where it's underused: large-population cardiovascular conditions like heart failure, where the biology is better understood but precision approaches haven't been fully applied.
AI in Drug Discovery: Integration Over Replacement
On artificial intelligence, Coloma was measured. Natural language processing tools and structure prediction platforms like AlphaFold have genuinely accelerated specific workflows — literature synthesis, protein structure prediction, certain aspects of computational drug design. These are real improvements.
But AI does not replace cryo-EM, X-ray crystallography, or experienced drug hunters. "Having proper drug hunters is important — people who know what a medicine looks like and can guide, in principle, whether something has a good shot at getting to a data set and maybe even to an approved product." The integration of AI into existing workflows is where the value is, not the replacement of deep experimental biology or medicinal chemistry expertise with models.
What You'll Learn
- Why most genetic associations fail to translate into medicines — and what variant functionalization does differently
- How Maze uncovered the pathogenic mechanism behind APOL1-mediated kidney disease, including the discovery of a novel protective variant
- Why APOL1 kidney disease presents earlier, progresses faster, and doesn't respond to standard-of-care therapies
- What the clinical journey looks like for a patient with APOL1-mediated kidney disease — from first presentation to dialysis
- How Maze is building the testing and advocacy infrastructure required for precision nephrology
- Why authentic community engagement looks different in a disease that disproportionately affects Black communities
- Which disease areas are suited to genetics-driven platforms — and which aren't yet
- Where AI meaningfully improves drug discovery workflows and where it doesn't replace experimental biology
Episode Highlights
- [00:00:46] From Genentech to Building Maze Therapeutics
- [00:04:21] Applying Precision Medicine Beyond Oncology
- [00:07:23] Moving Beyond Statistical Genetic Associations
- [00:11:32] APOL1 Mechanism Discovery
- [00:14:59] Clinical Reality of APOL1-Mediated Kidney Disease
- [00:20:15] Building Infrastructure for Genetic Testing
- [00:24:56] Authentic Community Engagement
- [00:27:47] Near-Term Milestones and Pipeline Progress
- [00:29:07] Where Genetics-Driven Platforms Work — and Where They Don't
- [00:35:46] AI in Drug Discovery: Integration Over Replacement
